fix: Narrowing for isinstance and type[T] is unsound

[tannguyencse19]

Summary

Fix unsound isinstance negative narrowing by skipping else-branch refinement when the classinfo is a runtime type[T] value rather than a literal class/ tuple/union of classes.

Fixes #713

Test Plan

cargo test -p pyrefly -- test_isinstance_type_classinfo_no_negative_narrow

[rchen152]

Is it possible to specifically disallow type[T] rather than trying to enumerate everything that is allowed? This implementation makes me a bit nervous, since it seems easy to miss cases that we should allow.

[tannguyencse19]

Let me check that

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

stone (https://github.com/dropbox/stone)
- ERROR stone/backends/python_rsrc/stone_base.py:105:75-90: Expected class object, got `Never` [invalid-argument]
- ERROR stone/backends/python_rsrc/stone_base.py:155:69-83: Expected class object, got `Never` [invalid-argument]

pandera (https://github.com/pandera-dev/pandera)
+ ERROR pandera/backends/ibis/base.py:168:17-36: Object of class `NoneType` has no attribute `rename` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:57:34-50: Object of class `object` has no attribute `sample` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:178:24-49: Object of class `NoneType` has no attribute `columns` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:181:40-70: Object of class `NoneType` has no attribute `with_columns` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:183:29-51: Object of class `NoneType` has no attribute `rows` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:187:40-64: Object of class `NoneType` has no attribute `rename` [missing-attribute]
+ ERROR pandera/backends/polars/base.py:188:26-54: Cannot index into `object` [bad-index]

porcupine (https://github.com/Akuli/porcupine)
- ERROR porcupine/plugins/highlight/pygments_highlighter.py:51:12-52: Object of class `type` has no attribute `get_tokens_unprocessed` [missing-attribute]

pytest-robotframework (https://github.com/detachhead/pytest-robotframework)
- ERROR pytest_robotframework/__init__.py:136:35-37: `TracebackType | None` is not assignable to attribute `__traceback__` with type `Never` [bad-assignment]

strawberry (https://github.com/strawberry-graphql/strawberry)
+ ERROR strawberry/http/__init__.py:28:26-39: Object of class `NoneType` has no attribute `errors` [missing-attribute]
+ ERROR strawberry/http/__init__.py:28:41-58: Object of class `NoneType` has no attribute `extensions` [missing-attribute]
+ ERROR strawberry/http/__init__.py:30:17-28: Object of class `NoneType` has no attribute `data` [missing-attribute]
+ ERROR strawberry/schema/schema.py:540:12-25: Object of class `NoneType` has no attribute `errors` [missing-attribute]
+ ERROR strawberry/schema/schema.py:541:44-57: `list[GraphQLError] | object | Unknown` is not assignable to attribute `pre_execution_errors` with type `list[GraphQLError] | None` [bad-assignment]
+ ERROR strawberry/schema/schema.py:543:38-51: Argument `list[GraphQLError] | object | Unknown` is not assignable to parameter `errors` with type `list[GraphQLError]` in function `strawberry.schema.base.BaseSchema._process_errors` [bad-argument-type]
+ ERROR strawberry/schema/schema.py:545:63-76: Argument `list[GraphQLError] | object | Unknown | None` is not assignable to parameter `errors` with type `list[GraphQLError] | None` in function `strawberry.types.execution.ExecutionResult.__init__` [bad-argument-type]
+ ERROR strawberry/schema/schema.py:546:9-26: Object of class `NoneType` has no attribute `extensions` [missing-attribute]
+ ERROR strawberry/schema/schema.py:548:16-22: Returned type `ExecutionResult | NoneType | Unknown` is not assignable to declared return type `ExecutionResult` [bad-return]

pylint (https://github.com/pycqa/pylint)
+ ERROR pylint/checkers/async_checker.py:74:25-41: Object of class `NoneType` has no attribute `getattr` [missing-attribute]
+ ERROR pylint/checkers/async_checker.py:75:25-41: Object of class `NoneType` has no attribute `getattr` [missing-attribute]
+ ERROR pylint/checkers/typecheck.py:1940:25-41: Object of class `NoneType` has no attribute `getattr` [missing-attribute]
+ ERROR pylint/checkers/typecheck.py:1941:25-41: Object of class `NoneType` has no attribute `getattr` [missing-attribute]
+ ERROR pylint/checkers/typecheck.py:1957:69-82: Object of class `NoneType` has no attribute `name` [missing-attribute]

pandas (https://github.com/pandas-dev/pandas)
+ ERROR pandas/core/apply.py:774:35-38: Cannot set item in `MutableMapping[Hashable, ((...) -> Unknown) | list[AggFuncTypeBase] | str]` [unsupported-operation]
+ ERROR pandas/core/arrays/arrow/array.py:598:24-45: Object of class `ExtensionArray` has no attribute `__arrow_array__`
+ Object of class `ndarray` has no attribute `__arrow_array__` [missing-attribute]
+ ERROR pandas/core/arrays/arrow/array.py:665:73-78: Argument `ExtensionArray | ndarray[tuple[Any, ...], dtype[Any]] | Any` is not assignable to parameter `values` with type `Collection[Unknown]` in function `pandas.core.dtypes.cast.construct_1d_object_array_from_listlike` [bad-argument-type]
- ERROR pandas/core/arrays/arrow/array.py:1209:16-60: Returned type `ArrowExtensionArray | Unknown` is not assignable to declared return type `NAType | bool` [bad-return]
+ ERROR pandas/core/arrays/arrow/array.py:1209:16-60: Returned type `ArrowExtensionArray | NAType | NaTType | Timedelta | Timestamp | Unknown` is not assignable to declared return type `NAType | bool` [bad-return]
- ERROR pandas/core/arrays/arrow/array.py:1273:16-60: Returned type `ArrowExtensionArray | Unknown` is not assignable to declared return type `NAType | bool` [bad-return]
+ ERROR pandas/core/arrays/arrow/array.py:1273:16-60: Returned type `ArrowExtensionArray | NAType | NaTType | Timedelta | Timestamp | Unknown` is not assignable to declared return type `NAType | bool` [bad-return]
+ ERROR pandas/core/arrays/string_arrow.py:521:24-37: Object of class `NAType` has no attribute `astype`
+ Object of class `NaTType` has no attribute `astype`
+ Object of class `Timedelta` has no attribute `astype`
+ Object of class `Timestamp` has no attribute `astype` [missing-attribute]
- ERROR pandas/core/construction.py:605:24-30: No matching overload found for function `list.__init__` called with arguments: (ndarray[tuple[Any, ...], dtype[Any]] | object) [no-matching-overload]
+ ERROR pandas/core/construction.py:605:24-30: No matching overload found for function `list.__init__` called with arguments: (ExtensionArray | ndarray[tuple[Any, ...], dtype[Any]] | object) [no-matching-overload]
- ERROR pandas/core/construction.py:658:20-26: No matching overload found for function `list.__init__` called with arguments: (object) [no-matching-overload]
+ ERROR pandas/core/construction.py:658:20-26: No matching overload found for function `list.__init__` called with arguments: (ExtensionArray | object) [no-matching-overload]
- ERROR pandas/core/ops/array_ops.py:129:48-57: Argument `ExtensionArray | ndarray[tuple[int], dtype[object_]] | ndarray[tuple[int], dtype[Any]]` is not assignable to parameter `right` with type `ndarray[tuple[Any, ...], dtype[object_]]` in function `pandas._libs.ops.vec_compare` [bad-argument-type]
+ ERROR pandas/core/ops/array_ops.py:129:48-57: Argument `ExtensionArray | Index | ndarray[tuple[int], dtype[object_]] | ndarray[tuple[int], dtype[Any]] | Unknown` is not assignable to parameter `right` with type `ndarray[tuple[Any, ...], dtype[object_]]` in function `pandas._libs.ops.vec_compare` [bad-argument-type]
- ERROR pandas/plotting/_matplotlib/core.py:443:16-19: Returned type `list[tuple[int | ndarray[tuple[Any, ...], dtype[Any]] | slice[Any, Any, Any]]]` is not assignable to declared return type `bool | list[tuple[int, ...]]` [bad-return]
+ ERROR pandas/plotting/_matplotlib/core.py:443:16-19: Returned type `list[tuple[int | ndarray[tuple[Any, ...], dtype[Any]] | slice[Any, Any, Any] | Unknown]]` is not assignable to declared return type `bool | list[tuple[int, ...]]` [bad-return]
+ ERROR pandas/tests/tseries/offsets/test_ticks.py:261:16-39: Object of class `Tick` has no attribute `_as_pd_timedelta` [missing-attribute]
+ ERROR pandas/tests/tseries/offsets/test_ticks.py:268:16-39: Object of class `Tick` has no attribute `_as_pd_timedelta` [missing-attribute]

static-frame (https://github.com/static-frame/static-frame)
+ ERROR static_frame/core/container_util.py:1332:22-33: Type `int` is not iterable [not-iterable]
+ ERROR static_frame/core/container_util.py:1771:49-60: Type `int` is not iterable [not-iterable]
+ ERROR static_frame/core/frame.py:2124:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:2170:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:2215:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:2256:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:2297:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:2329:30-39: Argument `StringIO | str` is not assignable to parameter `fp` with type `SupportsRead[bytes | str]` in function `json.load` [bad-argument-type]
+ ERROR static_frame/core/frame.py:4682:53-64: Type `int` is not iterable [not-iterable]
+ ERROR static_frame/core/frame.py:4684:67-87: `not in` is not supported between `int` and `int` [not-iterable]
+ ERROR static_frame/core/frame.py:6202:61-72: Type `int` is not iterable [not-iterable]
+ ERROR static_frame/core/index.py:941:31-44: No matching overload found for function `list.__init__` called with arguments: (int | list[int]) [no-matching-overload]
+ ERROR static_frame/core/index_hierarchy.py:1338:32-58: No matching overload found for function `map.__new__` called with arguments: (type[map[_S]], BoundMethod[Self@IndexHierarchy, (self: Self@IndexHierarchy, depth_level: TDepthLevelSpecifier = 0, /) -> ndarray[Any, Any]], int | list[int]) [no-matching-overload]
+ ERROR static_frame/core/index_hierarchy.py:1676:42-80: No matching overload found for function `zip.__new__` called with arguments: (type[zip[_T_co]], int | list[int], range) [no-matching-overload]
+ ERROR static_frame/core/index_hierarchy.py:1676:66-77: Argument `int | list[int]` is not assignable to parameter `obj` with type `Sized` in function `len` [bad-argument-type]
+ ERROR static_frame/core/index_hierarchy.py:1707:25-56: No matching overload found for function `map.__new__` called with arguments: (type[map[Index[Any]]], BoundMethod[list[Index[Any]], Overload[(self: list[Index[Any]], i: SupportsIndex, /) -> Index[Any], (self: list[Index[Any]], s: slice[Any, Any, Any], /) -> list[Index[Any]]]], int | list[int]) [no-matching-overload]
+ ERROR static_frame/core/index_hierarchy.py:1743:20-22: Argument `int | list[int]` is not assignable to parameter `obj` with type `Sized` in function `len` [bad-argument-type]
+ ERROR static_frame/core/index_hierarchy.py:1744:23-29: Object of class `int` has no attribute `pop` [missing-attribute]
+ ERROR static_frame/core/index_hierarchy.py:1858:24-28: No matching overload found for function `sorted` called with arguments: (int | list[int]) [no-matching-overload]
+ ERROR static_frame/core/index_hierarchy.py:2547:20-22: Argument `int | list[int]` is not assignable to parameter `obj` with type `Sized` in function `len` [bad-argument-type]
+ ERROR static_frame/core/index_hierarchy.py:2548:23-29: Object of class `int` has no attribute `pop` [missing-attribute]
+ ERROR static_frame/core/series.py:1154:25-33: `int | integer[Any] | list[int] | ndarray[Any, Any] | slice[Any, Any, Any] | None` is not assignable to variable `iloc_many` with type `list[int] | ndarray[Any, Any] | slice[Any, Any, Any] | None` [bad-assignment]
+ ERROR static_frame/core/type_blocks.py:1534:21-24: Argument `int | integer[Any] | list[int] | ndarray[Any, Any] | slice[Any, Any, Any]` is not assignable to parameter `__key` with type `list[int] | ndarray[tuple[Any, ...], dtype[Any]] | slice[Any, Any, Any]` in function `arraykit.BlockIndex.iter_contiguous` [bad-argument-type]
+ ERROR static_frame/core/yarn.py:498:21-28: Argument `int | ndarray[Any, dtype[signedinteger[_64Bit]]]` is not assignable to parameter `indexer` with type `ndarray[Any, dtype[signedinteger[_64Bit]]] | None` in function `Yarn.__init__` [bad-argument-type]

optuna (https://github.com/optuna/optuna)
+ ERROR tests/visualization_tests/test_slice.py:85:16-25: Cannot index into `Axes` [bad-index]

kopf (https://github.com/nolar/kopf)
+ ERROR kopf/_cogs/structs/dicts.py:244:19-23: Type of yielded value `(Iterable[Iterable[_T] | _T] & KubernetesModel) | (Iterable[_T] & KubernetesModel) | (KubernetesModel & _T)` is not assignable to declared return type `_T` [invalid-yield]